• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2010, Vol. 32 ›› Issue (1): 67-70.doi: 10.3969/j.issn.1007130X.2010.

• 论文 • 上一篇    下一篇

动摄像机下基于SIFT特征匹配和MHI的目标检测

  

  1. (国防科学技术大学电子科学与工程学院,湖南 长沙 410073)
  • 收稿日期:2008-08-28 修回日期:2008-11-17 出版日期:2010-01-18 发布日期:2010-01-18
  • 通讯作者: 410073 湖南省长沙市国防科学技术大学电子科学与工程学院智能感知信息联合中心 E-mail:wlf_0714@163.com
  • 作者简介:王亮芬(1982-),男,山东潍坊人,硕士生,研究方向为计算机视觉和图形图像处理;周东翔,副教授,研究方向为计算机视觉、虚拟现实、图形图像处理与科学计算可视化;梁华,博士生,研究方向为运动目标检测与跟踪;蔡宣平,教授,研究方向为计算机图形学、信息处理等。

The Detection of Moving Objects by a MovingCamera Based on the SIFT Features Match and MHI

  • Received:2008-08-28 Revised:2008-11-17 Online:2010-01-18 Published:2010-01-18

摘要:

摄像机的运动使得复杂背景下动目标的检测复杂化。为了应对动态变化的背景,本文提出基于SIFT特征匹配和运动历史图的目标检测算法。首先用SIFT算法提取特征点,采用RANSAC方法求得仿射变换模型参数并实现图像的全局运动补偿,最后利用运动历史图的方法检测出动目标。SIFT特征点匹配的准确性和RANSAC方法去除异常点的有效性使得仿射变换模型参数计算准确,运动历史图则给出了动目标清晰的轮廓,并指明了动目标的运动方向。与Ninad Thakoor实验结果对比说明:该算法能够准确地检测出动目标,并且显示了动目标的运动方向。

关键词: 运动摄像机, 全局运动补偿, SIFT, RANSAC, MHI, 目标检测

Abstract:

The movement of a camera makes the moving objects detection more difficult under a complex background. In order to cope with the dynamically changed background, we propose an object detection method based on the SIFT features match and Motion History Image. Firstly, the feature points are detected by the SIFT algorithm to compute the parameters of the affine transform model, guided by RANSAC, to compensate the global motion between images. Secondly, we adopt the MHI method to detect moving objects. The robustness of the SIFT features match and the validity of picking out the outliers by the  RANSAC algorithm make the parameters of the affine transform model compute accurately, and MHI shows the moving objects and the moving direction of objects clearly. The experimental results demonstrate that our algorithm can detect the moving objects accurately, and show the moving direction of the foreground objects, compared  with the Ninad Thakoor's method.

Key words: moving camera;global motion compensation;SIFT;RANSAC;MHI;objects detection

中图分类号: